Creating an element-wise minimum Series from two other Series in Python Pandas

Question

I am having trouble finding a way to do an efficient element-wise minimum of two Series objects in pandas. For example I can add two Series easily enough:

In [1]:
import pandas as pd
s1 = pd.Series(data=[1,1,1], index=[1,2,3])
s2 = pd.Series(data=[1,2,2,1], index=[1,2,3,4])
s1.add(s2)    
Out[1]:
1     2
2     3
3     3
4   NaN
dtype: float64

But I cannot find an efficient way to do an element-wise minimum between two Series (along with aligning the indices and handling NaN values).

Nevermind. There is an escape hatch with the combine function so you can put in any element-wise function:

In [2]:
s1 = pd.Series(data=[1,1,1], index=[1,2,3])
s2 = pd.Series(data=[1,2,2,1], index=[1,2,3,4])
s1.combine(s2, min, 0)
Out[2]:
1    1
2    1
3    1
4    0
dtype: int64

I think you should add the bit after "nevermind" as an answer, rather than at the end of the question (I'd upvote)... Although I disagree that `min(1, NaN) == 0`.... — Andy Hayden, Jun 07 '13 at 21:46
Yeah! You should answer your own question, so it has an accepted answer. — spen.smith, Apr 28 '19 at 23:04

score 61 · Answer 1 · edited Jan 09 '18 at 07:49

61

I find this the simplest:

import numpy as np

smax = np.minimum(s1, s2)

Link to docs (numpy.minimum)

edited Jan 09 '18 at 07:49

Jason

4,346
10
49
75

answered Sep 07 '17 at 05:01

Ankur Kanoria

957
8
9

14

Numpy ignores indexes, so this only works if the two series have the same index in the same order. – Imre Kerr Jan 24 '19 at 06:56
2

Short and sweet! – Zephaniah Grunschlag Mar 14 '21 at 06:08

score 53 · Answer 2 · answered Jun 07 '13 at 17:52

53

The most straightforward way I can see is to make them into a DataFrame and then take the row-wise min:

>>> print pandas.concat([s1, s2], axis=1).min(axis=1)
1    1
2    1
3    1
4    1
dtype: float64

answered Jun 07 '13 at 17:52

BrenBarn

242,874
37
412
384

2

On pandas 0.19.2 this is faster than other ways by more than three orders of magnitude – Eugene Pakhomov Jan 21 '17 at 18:36

score 28 · Answer 3 · answered Jun 07 '13 at 21:42

28

Another similar way:

In [11]: pd.DataFrame([s1, s2]).min()
Out[11]:
1    1
2    1
3    1
4    1
dtype: float64

answered Jun 07 '13 at 21:42

Andy Hayden

359,921
101
625
535

4

This only works for me on pandas 0.17 if I call `min(axis = 1)` – ihadanny Sep 19 '16 at 11:45
May I adjust the example to make it equal to https://stackoverflow.com/a/54807176/562769 ? – Martin Thoma Feb 21 '19 at 12:33

featuredpeow · Answer 4 · 2019-08-23T15:08:57.610

11

pd.Series([1,2,3,4,5]).clip(upper=pd.Series([6,5,4,3,2]))

will get you:

0    1
1    2
2    3
3    3
4    2
dtype: int64

edited Aug 23 '19 at 15:08

answered Aug 23 '19 at 15:02

featuredpeow

2,061
1
19
18

score 3 · Answer 5 · answered May 10 '20 at 16:35

You can concat the dataframes and take the minimum, specifying level=0:

>>> s1 = pd.Series(data=[1,1,1], index=[1,2,3])
>>> s2 = pd.Series(data=[1,2,2,1], index=[1,2,3,4])
>>> pd.concat([s1, s2]).min(level=0)
1    1
2    1
3    1
4    1
dtype: int64

This approach also works on dataframes.

score 2 · Answer 6 · answered Apr 20 '20 at 14:59

You can use the combine method of a DataFrame with np.minimum as the argument. np.minimum has special handling for NaN and complex NaNs.

Indeed, the pandas docs for combine uses the np.minimum function to illustrate a "true element-wise combine":

>>> df1 = pd.DataFrame({'A': [5, 0], 'B': [2, 4]})
>>> df2 = pd.DataFrame({'A': [1, 1], 'B': [3, 3]})
>>> df1.combine(df2, np.minimum)
   A  B
0  1  2
1  0  3

score 2 · Answer 7 · answered Apr 28 '20 at 15:04

2

This method will do the job:

import pandas as pd


def elementwise_min(x, y):
    x[x > y] = y
    return x


a = pd.Series([1, 2, 3])
b = pd.Series([0, 2, 4])
elementwise_min(a, b)

answered Apr 28 '20 at 15:04

renan-eccel

181
10

score 2 · Answer 8 · answered Dec 12 '22 at 13:06

2

You can use clip method:

>>> s1.clip(upper=s2)
1    1
2    1
3    1
dtype: int64

>>> s2.clip(upper=s1)
1    1.0
2    1.0
3    1.0
4    NaN
dtype: float64

answered Dec 12 '22 at 13:06

Bhindi

1,403
12
16

score 1 · Answer 9 · answered Feb 21 '19 at 12:33

1

Same answer as Andy Hayden, but a bit easier to read:

>>> import pandas as pd
>>> s1 = pd.Series(data=[1,2,3,4,5], index=[1,2,3,4,5])
>>> s2 = pd.Series(data=[5,1,3,5], index=[1,2,3,4])
>>> pd.DataFrame([s1, s2]).min()
1    1.0
2    1.0
3    3.0
4    4.0
5    5.0
dtype: float64

answered Feb 21 '19 at 12:33

Martin Thoma

124,992
159
614
958

1

It's better to help edit Andy's answer to make it easier to read, than to submit a duplicate answer. – Ken Williams Dec 19 '19 at 16:43
1

@KenWilliams What is easy to read is personal preference (at least in this case). – Martin Thoma Dec 19 '19 at 16:52

Maxim Egorushkin · Answer 10 · 2022-05-17T05:14:32.923

1

One straight-forward way is to use Series.where function:

min_s1_s2 = s1.where(s1 < s2, s2)

where is a fundamental method in terms of which clip can be implemented, but not the other way around.

edited May 17 '22 at 05:14

answered May 16 '22 at 14:56

Maxim Egorushkin

131,725
17
180
271

Creating an element-wise minimum Series from two other Series in Python Pandas

10 Answers10

Linked

Related